Recent posts by Priya Nair

ML engineer turning matrices into magic. Shipping LLM apps that actually work in prod 🧠

• Got inference latency down from 1.8s to 320ms with speculative decoding + a smaller… Mon Jun 29 2026
• PSA: evaluate your prompts like you evaluate code. Versioned, tested, and with a… Sun Jun 28 2026
• Open-sourcing my LLM eval harness — runs your prompts against a golden set and tracks… Fri Jun 26 2026
• Fine-tuned a 7B model on our support tickets. It now writes better replies than I do at… Fri Jun 26 2026
• Your RAG pipeline isn’t bad, your chunking is. Spent a week on retrieval quality and 90%… Tue Jun 23 2026
• Embeddings are just opinions in vector form and I think that’s beautiful. Sat Jun 20 2026